Why Baby Talk Sounds Similar Around the World
A globetrotting study that examined how people talk to young infants across 21 different societies has found consistencies across cultures that suggest human “baby talk” may have evolved in response to intense functional pressure. The research is published in Nature Human Behaviour.
The features of baby talk
From the meerkat’s bark to the cheetah’s chirp (no, really) many vocalizations in the animal kingdom often have a clear function – for example, signaling nearby predators or safety. These functions can limit and direct the form that these noises have – an alarm call must necessarily be loud and distinct, for example.
Are the common vocalization humans make also shaped by their purpose? A 40-strong team of international researchers have now conducted an investigation into the form and function of baby talk. Previous work in this area, like many lab studies, had been weakened by relying too much on educated subsets of Western societies. To remedy this, the team sampled from 21 human societies, producing a dataset that covered 18 different languages across 12 language families. The team amassed 1,615 recordings of baby- and adult-directed speech and song. The populations analyzed featured four societies, including the Hadza people of northern Tanzania, who remain relatively isolated from global media.
The study analyzed 15 different acoustic features of the baby talk recordings – including rhythm and pitch – to create a dataset that could be fed into a machine learning algorithm. The team used an approach called k-fold cross-validation, which involved training the algorithm on 20 out of 21 language datasets and then testing on the remaining language to determine whether the algorithm could tell baby talk from adult-directed speech. The “test” language was then rotated out for another culture’s recordings – a process repeated 20 times to produce a similarity value for the sample sets.
The team found that, for all the cultures, both speech and song could be predicted by the model as being infant- or adult-directed at a level higher than if the algorithm had been guessing at random, suggesting that the different cultures’ baby-talk used common acoustic features. The similarity, for speech, was strongly driven by pitch.
50,000 listeners lend an ear
To determine if human listeners could also identify baby talk and lullabies, the team played vocalizations to a vast group of 51,065 people who took part via a citizen science platform called The Music Lab. The listeners, from 187 countries, could guess whether a vocalization was directed and a baby or adult at a level significantly above chance. Recordings from some locations were easier to discern than others – listeners found baby talk produced by people from Wellington, New Zealand, the easiest to identify, but struggled with similar recordings of the Bislama language of Vanuatu.
The authors write, “despite evident variability in language, music and infant care practices worldwide, when people speak or sing to fussy infants, they modify the acoustic features of their vocalizations in similar and mutually intelligible ways across cultures” in the publication. The team also noted that some features of infant-directed speech and song, such as a less harsh timbre and higher pitch, are similarly found in “friendly” vocalizations across the animal kingdom, suggesting a common evolutionary mechanism might be at work.
Reference: Hilton CB, Moser CJ, Bertolo M et al. Acoustic regularities in infant-directed speech and song across cultures. Nat. Hum. Behav. 2022. doi: 10.1038/s41562-022-01410-x